Edge computing driven partial discharge real-time monitoring and diagnosis system and method

Abstract

The application discloses an edge-computing-driven partial discharge real-time monitoring and diagnosis system and method, relates to the technical field of intelligent diagnosis of partial discharge, and comprises a partial discharge monitoring terminal plane moving device, a plurality of equipment partial discharge edge monitoring terminals and a central server. The application is characterized in that a plurality of plane rails are installed with the partial discharge monitoring terminal plane moving device, a plane electric sliding block is located on the corresponding plane rail to move on the plane, can cover the area above the plane rail, and can realize the matching of the monitoring positions of the to-be-monitored equipment in the area range under the joint action of the electric lifting mechanism, so that the shielding in the lateral monitoring can be avoided as much as possible to cause the monitoring failure. Through the organic combination of edge intelligence and mobile monitoring, the change from the regular sampling inspection to the continuous panoramic monitoring of the power equipment state monitoring is realized, and the application is especially suitable for the equipment health management of complex scenes such as transformer substations and offshore wind power.

Description

Technical Field

[0001] This invention relates to the field of intelligent partial discharge diagnosis technology, and in particular to a real-time monitoring and diagnosis system and method for partial discharge driven by edge computing. Background Technology

[0002] Partial discharge (PD), as a precursor to insulation faults in power equipment, plays a crucial early warning role in power systems. With the development of power systems towards higher voltage, larger capacity, and greater intelligence, real-time partial discharge monitoring technology is playing an increasingly important role in ensuring safe equipment operation and improving power supply reliability. However, traditional partial discharge monitoring technologies face numerous challenges, including weak signals, significant noise interference, and complex data processing.

[0003] Partial discharge real-time monitoring technology is an important means of assessing the condition of power equipment. Its core lies in capturing partial discharge signals generated during the operation of power equipment in real time through various sensors, and identifying the discharge type and location through signal processing and analysis technology, so as to ultimately achieve the assessment of the insulation status of the equipment and the early warning of faults.

[0004] Partial discharge signals are typically weak and highly susceptible to environmental noise and electromagnetic interference. This problem is particularly prominent in the partial discharge monitoring of high-voltage electrical equipment, making it extremely difficult to accurately capture and analyze partial discharge signals.

[0005] In practical applications, various electromagnetic interference sources exist around power equipment, such as other electrical equipment, radio transmissions, and even lightning activity. These interferences can mask or distort partial discharge signals, leading to inaccurate monitoring results. Furthermore, the partial discharge signal itself has low energy and is prone to attenuation during propagation, further increasing the difficulty of monitoring.

[0006] Real-time monitoring of partial discharge generates a massive amount of data. Effectively processing and analyzing this data to extract useful information is a significant challenge. With the surge in the number of power equipment, especially distribution network equipment, the sheer volume of partial discharge monitoring data exacerbates the burden on the system's computation, transmission, and storage.

[0007] Traditional data processing methods often require transmitting all data to a central server for processing. This not only consumes a large amount of bandwidth but can also lead to processing delays, affecting the real-time performance of monitoring. Furthermore, the analysis of partial discharge data requires specialized knowledge and experience. Developing accurate and reliable analysis algorithms is both a key focus and a challenge in current research. Summary of the Invention

[0008] To address the aforementioned technical problems, this invention provides an edge computing-driven real-time monitoring and diagnosis system and method for partial discharge. The technical solution adopted is as follows:

[0009] The edge computing-driven real-time partial discharge monitoring and diagnosis system includes a partial discharge monitoring terminal planar moving device, multiple device partial discharge edge monitoring terminals, and a central server. The multiple device partial discharge edge monitoring terminals are detachably installed on the bottom of multiple planar electric sliders of the partial discharge monitoring terminal planar moving device, and move locally on top of multiple devices under test following the multiple planar electric sliders. The device partial discharge edge monitoring terminals monitor the partial discharge signals and visual data of the devices under test, and analyze whether the devices under test have a partial discharge state based on the partial discharge signals and visual data. If a partial discharge state is determined to exist, the terminal interacts with the central server to exchange the corresponding partial discharge signals and visual data. If the device partial discharge edge monitoring terminal determines that no partial discharge state exists, the terminal stores the partial discharge signals and visual data at the edge for future reference.

[0010] Optionally, the partial discharge monitoring terminal planar movement device includes multiple planar tracks, multiple planar electric sliders, and multiple electric lifting mechanisms. The multiple planar tracks are respectively installed on the top of the room where multiple devices to be tested are located. The multiple planar electric sliders are respectively installed on the multiple planar tracks. Each planar electric slider moves locally within the area of ​​its corresponding planar track. The bases of the multiple electric lifting mechanisms are respectively installed on the multiple planar electric sliders. The lifting parts of the multiple electric lifting mechanisms are vertically downward. The multiple device partial discharge edge monitoring terminals are detachably installed at the ends of the lifting parts of the electric lifting mechanisms. The device partial discharge edge monitoring terminals move within the space above the devices to be tested under the driving action of the planar electric sliders and electric lifting mechanisms.

[0011] By adopting the above technical solution, multiple planar tracks are installed with the partial discharge monitoring terminal planar moving device. The planar electric slider moves locally on the corresponding planar track, which can cover the area above the planar track. Multiple equipment partial discharge edge monitoring terminals are detachably installed at the lifting end of the electric lifting mechanism. Under the joint action of the electric lifting mechanism, the monitoring position of the equipment to be monitored within the area can be matched. The equipment partial discharge edge monitoring terminal is located directly above the equipment to be monitored, which can achieve more accurate partial discharge signal and visual data acquisition. This acquisition method can avoid the situation of monitoring failure caused by obstruction in lateral monitoring as much as possible.

[0012] The equipment partial discharge edge monitoring terminal uses edge processing to process the corresponding partial discharge signals and visual data of the equipment to be monitored below. In most cases, when no partial discharge state occurs, the edge calculation and judgment of the state can significantly reduce the data processing volume of the central server. Only when the equipment partial discharge edge monitoring terminal determines that a partial discharge state has occurred will the corresponding partial discharge signals and visual data be exchanged.

[0013] By organically combining edge intelligence and mobile monitoring, the status monitoring of power equipment has been transformed from "periodic spot checks" to "continuous panoramic monitoring". It is particularly suitable for equipment health management in complex scenarios such as substations and offshore wind power, and can significantly improve the early warning of equipment failures.

[0014] Optionally, the device partial discharge edge monitoring terminal includes a housing, a multimodal sensor module, a visible light visual monitoring module, an edge computing module, a positioning module, a memory, and an edge communication module. One side of the housing is detachably mounted on the lifting end of the electric lifting mechanism. The multimodal sensor module and the visible light visual monitoring module are respectively installed inside the housing, with the sensor head located on the other side of the housing facing the device under test. The multimodal sensor module monitors pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, and transient voltage signals. The visible light visual monitoring module monitors the visual image. The data output terminals of the multimodal sensor module and the visible light visual monitoring module are respectively connected to the edge computing module. The edge computing module calculates whether a partial discharge state exists based on the pulse current signal, mechanical vibration wave signal, electromagnetic wave signal, transient voltage signal, and visual image. The positioning module collects the current location data of the device partial discharge edge monitoring terminal. The multimodal sensor module, the visible light visual monitoring module, the edge computing module, and the positioning module are respectively connected to the memory. The memory is connected to the central server through the edge communication module.

[0015] Optionally, the multimodal sensor module includes a high-frequency current sensor, an ultrasonic sensor, an ultra-high frequency sensor, and a transient voltage sensor. The signal output terminals of the high-frequency current sensor, ultrasonic sensor, ultra-high frequency sensor, and transient voltage sensor are respectively connected to the edge computing module for edge communication, installed in the housing, and the sensing head faces the device to be detected.

[0016] Optionally, the visible light visual monitoring module includes a housing, a visual camera, and a thermal infrared sensor. The housing is installed inside the housing and faces the device to be inspected. The visual camera and the thermal infrared sensor are installed inside the housing and located in the middle of the housing. The sensing heads of the visual camera and the thermal infrared sensor face the device to be inspected and are respectively connected to the edge computing module.

[0017] By adopting the above technical solutions, the multimodal full-domain sensing capability is improved. It integrates four types of sensors: high-frequency current (>MHz), ultrasonic (40-300kHz), ultra-high frequency (300MHz-3GHz), and transient voltage (ns-level resolution), covering the current, mechanical vibration, electromagnetic radiation, and voltage transient characteristics of partial discharge, realizing the full-dimensional synchronous acquisition of discharge signals and significantly reducing the missed detection rate.

[0018] The visible light camera and the thermal infrared sensor work together to simultaneously capture surface deformation, discharge spots and abnormal temperature rise of the equipment, supporting the identification of micro-defects at the 0.1mm level, and greatly improving the overall fault feature matching degree.

[0019] Edge computing modules integrate lightweight neural networks (such as MobileNet-SSD), supporting spatiotemporal alignment and fusion of multi-sensor signals and visual data, significantly reducing the response time of diagnostic algorithms and greatly improving efficiency compared to traditional single-modal analysis.

[0020] It integrates a millimeter-level positioning module (UWB / laser SLAM), which supports ±1mm positioning accuracy of the monitoring terminal on the electric lifting mechanism, and is compatible with equipment of different sizes (such as GIS bays and transformer bushings), thus shortening the deployment time.

[0021] By deeply integrating multimodal perception and edge intelligence, a closed-loop management system of "holographic perception - precise positioning - intelligent judgment - rapid response" for partial discharge of power equipment has been realized. It is especially suitable for scenarios with high reliability requirements such as new energy power plants and ultra-high voltage converter stations, which can significantly improve the average early warning time of equipment insulation faults and significantly reduce the overall operation and maintenance costs.

[0022] Optionally, the edge computing module includes an edge memory, an edge data analysis chip, and a vision analysis chip. The edge memory is communicatively connected to a high-frequency current sensor, an ultrasonic sensor, an ultra-high frequency sensor, a transient voltage sensor, a vision camera, and a thermal infrared sensor, respectively, to collect pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, transient voltage signals, and visual images. The edge data analysis chip and the vision analysis chip are communicatively connected to the edge memory and the vision analysis chip, respectively, to calculate whether a partial discharge state has occurred based on sensor data and visual analysis results. The edge data analysis chip is communicatively connected to the central server through an edge communication module.

[0023] By adopting the above technical solutions, the edge memory can store pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, transient voltage signals and visual images at the edge for subsequent retrieval. The edge data analysis chip can extract electrical features and perform edge calculations based on multimodal data to determine the probability of partial discharge states. The visual analysis chip can efficiently process visual image data.

[0024] Optionally, the central server includes a central communication module and a computer, wherein the computer is connected to the edge communication module through the central communication module.

[0025] An edge computing-driven real-time monitoring and diagnosis method for partial discharge uses an edge computing-driven real-time monitoring and diagnosis system to monitor and diagnose partial discharge in multiple devices under test indoors, including the following steps:

[0026] Step 1: Control the planar electric slider of the corresponding area above each device to move along the planar track to directly above the device to be tested, and drive the electric lifting mechanism to lower the device partial discharge edge monitoring terminal to a set distance from the device;

[0027] Step 2: The edge memory communicates with the high-frequency current sensor, ultrasonic sensor, ultra-high frequency sensor, transient voltage sensor, vision camera and thermal infrared sensor to collect pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, transient voltage signals and visual images, and preprocesses the multimodal data respectively.

[0028] Step 3: The edge data analysis chip extracts electrical signal features, including pulse current features, ultrasonic energy density, ultra-high frequency signal strength, and transient voltage change rate.

[0029] Step 4: The visual analysis chip extracts visual features, including hotspot features and discharge spot features.

[0030] Step 5: The edge data analysis chip uses a weighted decision fusion algorithm to calculate the probability P of partial discharge state occurrence based on electrical signal features and visual features, and sets a threshold for the probability of partial discharge state occurrence. If judge If a partial discharge state is detected, the edge data analysis chip communicates with the central server via the edge communication module to exchange the partial discharge signal and visual data corresponding to the partial discharge state.

[0031] By adopting the above technical solution, the extraction of electrical signal characteristics can be performed using the following process, and the pulse current characteristics can be calculated using the following formula:

[0032] ; ;

[0033] in It is the peak current. For pulse current signals Take the maximum value. It is the pulse characteristic repetition rate;

[0034] Ultrasonic energy density The calculation formula is:

[0035] ;

[0036] in It is a mechanical vibration wave signal;

[0037] UHF signal strength The calculation formula is:

[0038] ;

[0039] in It is an electromagnetic wave signal;

[0040] Transient voltage change rate The calculation formula is:

[0041] ;

[0042] In visual feature extraction, hotspot features The calculation method is as follows:

[0043] ,

[0044] in It is a temperature matrix of thermal infrared data. It is the average indoor temperature;

[0045] The discharge spot feature uses a convolutional neural network (CNN) to monitor whether a discharge spot appears in an image and outputs a probability. .

[0046] Optionally, in step 1, the distance is set to be mm-500mm from the surface of the device to be monitored.

[0047] By adopting the above technical solution, the distance from the surface of the device to be monitored is kept between mm and 500mm, which avoids interference with the normal operation of the device and also minimizes interference with the monitoring signal caused by obstruction.

[0048] Optionally, in step 5, the formula for calculating the probability P of the partial discharge state is:

[0049] ;

[0050] in These are weighting coefficients. These are the normalized electrical characteristic values. It is the detection probability of the discharge spot characteristics;

[0051] Detection probability of discharge spot features The calculation formula is:

[0052] ;

[0053] in It is a visible light image frame. Represents the CNN model. It is a thermal infrared temperature matrix. yes The function maps the CNN output to the interval [0, 1], representing the confidence level of the existence of the discharge spot.

[0054] By adopting the above technical solution, if any of the following conditions are met, it is determined to be partial discharge: ,like When both electrical and visual characteristics exceed the threshold, such as... and After receiving abnormal data, the central server can use a deep residual network (ResNet) to jointly analyze the multimodal data and output the fault type (such as insulator cracks or conductor burrs).

[0055] In summary, the present invention has at least one of the following beneficial technical effects:

[0056] This invention provides an edge computing-driven real-time monitoring and diagnosis system and method for partial discharge. Multiple planar tracks are used to install partial discharge monitoring terminal planar movement devices. Planar electric sliders move locally along the corresponding planar tracks, covering the area above the track. Multiple device partial discharge edge monitoring terminals are detachably installed at the lifting end of an electric lifting mechanism. Under the combined action of the electric lifting mechanism, the monitoring positions of the devices to be monitored within the area can be matched. With the device partial discharge edge monitoring terminal positioned directly above the device to be monitored, more accurate partial discharge signals and visual data acquisition can be achieved. This acquisition method minimizes the risk of monitoring failure due to obstruction in lateral monitoring.

[0057] The equipment partial discharge edge monitoring terminal uses edge processing to process the corresponding partial discharge signals and visual data of the equipment to be monitored below. In most cases, when no partial discharge state occurs, the edge calculation and judgment of the state can significantly reduce the data processing volume of the central server. Only when the equipment partial discharge edge monitoring terminal determines that a partial discharge state has occurred will the corresponding partial discharge signals and visual data be exchanged.

[0058] By organically combining edge intelligence and mobile monitoring, the status monitoring of power equipment has been transformed from "periodic spot checks" to "continuous panoramic monitoring". It is particularly suitable for equipment health management in complex scenarios such as substations and offshore wind power, and can significantly improve the early warning of equipment failures. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the structural principle of the edge computing-driven real-time monitoring and diagnosis system for partial discharge of the present invention;

[0060] Figure 2 This is a schematic diagram of the electrical component connection principle of the edge computing-driven partial discharge real-time monitoring and diagnosis system of the present invention.

[0061] Explanation of reference numerals in the attached drawings: 1. Partial discharge monitoring terminal planar moving device; 11. Planar track; 2. Planar electric slider; 3. Equipment partial discharge edge monitoring terminal; 32. Multimodal sensor module; 321. High-frequency current sensor; 322. Ultrasonic sensor; 323. UHF sensor; 324. Transient voltage sensor; 33. Visible light visual monitoring module; 332. Visual camera; 333. Thermal infrared sensor; 34. Edge computing module; 341. Edge-end memory; 342. Edge-end data analysis chip; 343. Visual analysis chip; 35. Positioning module; 36. Memory; 37. Edge-end communication module; 4. Central server; 41. Central-end communication module; 42. Computer; 5. Electric lifting mechanism; 100. Indoor; 101. Equipment under test; Detailed Implementation

[0062] The present invention will be further described in detail below with reference to the accompanying drawings.

[0063] This invention discloses an edge computing-driven real-time monitoring and diagnosis system and method for partial discharge.

[0064] Reference Figure 1 and Figure 2 Example 1: A partial discharge real-time monitoring and diagnosis system driven by edge computing includes a partial discharge monitoring terminal planar moving device 1, multiple device partial discharge edge monitoring terminals 3, and a central server 4. The multiple device partial discharge edge monitoring terminals 3 are detachably installed at the bottom of multiple planar electric sliders 2 of the partial discharge monitoring terminal planar moving device 1, and move locally on the top of multiple devices 101 to be tested, following the multiple planar electric sliders 2. The device partial discharge edge monitoring terminals 3 monitor the partial discharge signals and visual data of the devices 101 to be tested, and analyze whether the devices 101 to be tested are in a partial discharge state based on the partial discharge signals and visual data. If it is determined that a partial discharge state exists, it interacts with the central server 4 to exchange the corresponding partial discharge signals and visual data. If the device partial discharge edge monitoring terminals 3 determine that no partial discharge state exists, it stores the partial discharge signals and visual data at the edge for future reference.

[0065] Example 2: The partial discharge monitoring terminal planar movement device 1 includes multiple planar tracks 11, multiple planar electric sliders 2, and multiple electric lifting mechanisms 5. The multiple planar tracks 11 are respectively installed on the top of the indoor 100 where multiple devices 101 to be tested are located. The multiple planar electric sliders 2 are respectively installed on the multiple planar tracks 11. Each planar electric slider 2 moves locally within the area of ​​the corresponding planar track 11. The bases of the multiple electric lifting mechanisms 5 are respectively installed on the multiple planar electric sliders 2. The lifting parts of the multiple electric lifting mechanisms 5 are vertically downward. Multiple device partial discharge edge monitoring terminals 3 are detachably installed at the ends of the lifting parts of the electric lifting mechanisms 5. The device partial discharge edge monitoring terminals 3 move within the top space of the device 101 to be tested under the driving action of the planar electric sliders 2 and the electric lifting mechanisms 5.

[0066] Multiple planar tracks 11 are installed on the partial discharge monitoring terminal planar moving device 1. The planar electric slider 2 moves locally on the corresponding planar track 11, which can cover the area above the planar track 11. Multiple device partial discharge edge monitoring terminals 3 are detachably installed at the lifting end of the electric lifting mechanism 5. Under the joint action of the electric lifting mechanism 5, the monitoring position of the device to be monitored 101 within the area can be matched. The device partial discharge edge monitoring terminal 3 is located directly above the device to be monitored 101, which can realize more accurate partial discharge signal and visual data acquisition. This acquisition method can avoid the situation of monitoring failure caused by obstruction in lateral monitoring as much as possible.

[0067] The equipment partial discharge edge monitoring terminal 3 uses edge processing to process the corresponding partial discharge signal and visual data of the device 101 to be monitored below. In most cases, when no partial discharge state occurs, the edge calculation and judgment of the state can significantly reduce the data processing volume of the central server 4. Only when the equipment partial discharge edge monitoring terminal 3 determines that a partial discharge state has occurred will the corresponding partial discharge signal and visual data be exchanged.

[0068] By organically combining edge intelligence and mobile monitoring, the status monitoring of power equipment has been transformed from "periodic spot checks" to "continuous panoramic monitoring". It is particularly suitable for equipment health management in complex scenarios such as substations and offshore wind power, and can significantly improve the early warning of equipment failures.

[0069] Example 3: The device partial discharge edge monitoring terminal 3 includes a housing, a multimodal sensor module 32, a visible light visual monitoring module 33, an edge computing module 34, a positioning module 35, a memory 36, and an edge communication module 37. One side of the housing is detachably mounted on the lifting end of the electric lifting mechanism 5. The multimodal sensor module 32 and the visible light visual monitoring module 33 are respectively installed inside the housing, and the sensor head is located on the other side of the housing facing the device 101 under test. The multimodal sensor module 32 monitors pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, and transient voltage signals, while the visible light visual monitoring module 33 monitors the visible light... The data output terminals of the multimodal sensor module 32 and the visible light visual monitoring module 33 are respectively connected to the edge computing module 34. The edge computing module 34 calculates whether there is a partial discharge state based on the pulse current signal, mechanical vibration wave signal, electromagnetic wave signal, transient voltage signal and visual image. The positioning module 35 collects the position data of the current device partial discharge edge monitoring terminal 3. The multimodal sensor module 32, the visible light visual monitoring module 33, the edge computing module 34 and the positioning module 35 are respectively connected to the memory 36. The memory 36 is connected to the central server 4 through the edge communication module 37.

[0070] Example 4: The multimodal sensor module 32 includes a high-frequency current sensor 321, an ultrasonic sensor 322, an ultra-high frequency sensor 323, and a transient voltage sensor 324. The signal output terminals of the high-frequency current sensor 321, ultrasonic sensor 322, ultra-high frequency sensor 323, and transient voltage sensor 324 are respectively connected to the edge computing module 34 for edge communication. They are installed in the housing, and the sensing heads face the device to be tested 101.

[0071] Example 5: The visible light visual monitoring module 33 includes a housing, a visual camera 332, and a thermal infrared sensor 333. The housing is installed inside the housing and faces the device to be tested 101. The visual camera 332 and the thermal infrared sensor 333 are installed inside the housing and located in the middle of the housing. The sensing heads of the visual camera 332 and the thermal infrared sensor 333 face the device to be tested 101 and are respectively connected to the edge computing module 34.

[0072] The multimodal full-domain sensing capability is enhanced by integrating four types of sensors: high-frequency current (>100MHz), ultrasonic (40-300kHz), ultra-high frequency (300MHz-3GHz), and transient voltage (ns-level resolution). This covers the current, mechanical vibration, electromagnetic radiation, and voltage transient characteristics of partial discharge, enabling synchronous acquisition of discharge signals across all dimensions and significantly reducing the missed detection rate.

[0073] The visible light camera and the thermal infrared sensor work together to simultaneously capture surface deformation, discharge spots and abnormal temperature rise of the equipment, supporting the identification of micro-defects at the 0.1mm level, and greatly improving the overall fault feature matching degree.

[0074] Edge computing modules integrate lightweight neural networks (such as MobileNet-SSD), supporting spatiotemporal alignment and fusion of multi-sensor signals and visual data, significantly reducing the response time of diagnostic algorithms and greatly improving efficiency compared to traditional single-modal analysis.

[0075] The integrated millimeter-level positioning module 35 (UWB / laser SLAM) supports ±1mm positioning accuracy of the monitoring terminal on the electric lifting mechanism, adapts to equipment of different sizes (such as GIS bays, transformer bushings), and shortens deployment time.

[0076] By deeply integrating multimodal perception and edge intelligence, a closed-loop management system of "holographic perception - precise positioning - intelligent judgment - rapid response" for partial discharge of power equipment has been realized. It is especially suitable for scenarios with high reliability requirements such as new energy power plants and ultra-high voltage converter stations, which can significantly improve the average early warning time of equipment insulation faults and significantly reduce the overall operation and maintenance costs.

[0077] Example 6: The edge computing module 34 includes an edge memory 341, an edge data analysis chip 342, and a vision analysis chip 343. The edge memory 341 is communicatively connected to a high-frequency current sensor 321, an ultrasonic sensor 322, an ultra-high frequency sensor 323, a transient voltage sensor 324, a vision camera 332, and a thermal infrared sensor 333, respectively, to collect pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, transient voltage signals, and visual images. The edge data analysis chip 342 and the vision analysis chip 343 are communicatively connected to the edge memory 341, respectively. The edge data analysis chip 342 is communicatively connected to the vision analysis chip 343 to calculate whether a partial discharge state has occurred based on sensor data and visual analysis results. The edge data analysis chip 342 is communicatively connected to the central server 4 through the edge communication module 37.

[0078] The edge memory 341 can store pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, transient voltage signals and visual images at the edge for later retrieval. The edge data analysis chip 342 can extract electrical features and perform edge calculations of the probability of partial discharge state occurrence based on multimodal data. The visual analysis chip 343 can efficiently process visual image data.

[0079] In Example 7, the central server 4 includes a central communication module 41 and a computer 42. The computer 42 is connected to the edge communication module 37 through the central communication module 41.

[0080] Example 8: A method for real-time monitoring and diagnosis of partial discharge driven by edge computing. This method uses an edge computing-driven real-time monitoring and diagnosis system to monitor and diagnose partial discharge in multiple devices 101 within an indoor space 100. The method includes the following steps:

[0081] Step 1: Control the planar electric slider 2 of the corresponding area above each device 101 to move along the planar track 11 to directly above the device 101, and drive the electric lifting mechanism 5 to lower the device partial discharge edge monitoring terminal 3 to a set distance from the device;

[0082] Step 2: The edge memory 341 communicates with the high-frequency current sensor 321, ultrasonic sensor 322, ultra-high frequency sensor 323, transient voltage sensor 324, vision camera 332 and thermal infrared sensor 333 respectively to collect pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, transient voltage signals and visual images, and preprocesses the multimodal data respectively.

[0083] Step 3: The edge data analysis chip 342 extracts electrical signal features, including pulse current features, ultrasonic energy density, ultra-high frequency signal strength, and transient voltage change rate.

[0084] Step 4: The visual analysis chip 343 extracts visual features, including hotspot features and discharge spot features.

[0085] Step 5: The edge data analysis chip 342 uses a weighted decision fusion algorithm to calculate the probability P of partial discharge state occurrence based on electrical signal features and visual features, and sets a threshold for the probability of partial discharge state occurrence. If judge If a partial discharge state is detected, the edge data analysis chip 342 communicates with the central server 4 through the edge communication module 37 to exchange the partial discharge signal and visual data corresponding to the partial discharge state.

[0086] The extraction of electrical signal features can be performed using the following process, and the pulse current features can be calculated using the following formula:

[0087] ; ;

[0088] in It is the peak current. For pulse current signals Take the maximum value. It is the pulse characteristic repetition rate;

[0089] Ultrasonic energy density The calculation formula is:

[0090] ;

[0091] in It is a mechanical vibration wave signal;

[0092] UHF signal strength The calculation formula is:

[0093] ;

[0094] in It is an electromagnetic wave signal;

[0095] Transient voltage change rate The calculation formula is:

[0096] ;

[0097] In visual feature extraction, hotspot features The calculation method is as follows:

[0098] ,

[0099] in It is a temperature matrix of thermal infrared data. It is the average indoor temperature;

[0100] The discharge spot feature uses a convolutional neural network (CNN) to monitor whether a discharge spot appears in an image and outputs a probability. .

[0101] In Example 9, in step 1, the set distance is 100mm-500mm from the surface of the device to be monitored 101.

[0102] The distance between the device to be monitored and the surface of the device 101 is 100mm-500mm to avoid interference with the normal operation of the device 101 and to minimize interference with the monitoring signal caused by obstruction.

[0103] In Example 10, step 5, the formula for calculating the probability P of the partial discharge state is:

[0104] ;

[0105] in These are weighting coefficients. These are the normalized electrical characteristic values. It is the detection probability of the discharge spot characteristics;

[0106] Detection probability of discharge spot features The calculation formula is:

[0107] ;

[0108] in It is a visible light image frame. Represents the CNN model. It is a thermal infrared temperature matrix. yes The function maps the CNN output to the interval [0, 1], representing the confidence level of the existence of the discharge spot.

[0109] A partial discharge is determined if any of the following conditions are met: ,like When both electrical and visual characteristics exceed the threshold, such as... and After receiving abnormal data, the central server 4 can use a deep residual network (ResNet) to jointly analyze the multimodal data and output the fault type (such as insulator cracks or conductor burrs).

[0110] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for real-time monitoring and diagnosis of partial discharge driven by edge computing, characterized in that: A partial discharge real-time monitoring and diagnostic system driven by edge computing is used to monitor and diagnose partial discharge in multiple devices (101) under test in an indoor environment (100), including the following steps: Step 1: Control the planar electric slider (2) of the corresponding area above each device to be tested (101) to move along the planar track (11) to directly above the device to be tested (101), and drive the electric lifting mechanism (5) to lower the device partial discharge edge monitoring terminal (3) to a set distance from the device; Step 2: The edge memory (341) communicates with the high-frequency current sensor (321), ultrasonic sensor (322), ultra-high frequency sensor (323), transient voltage sensor (324), vision camera (332) and thermal infrared sensor (333) respectively to collect pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, transient voltage signals and visual images, and preprocesses the multimodal data respectively; Step 3: The edge data analysis chip (342) extracts electrical signal features, including pulse current features, ultrasonic energy density, ultra-high frequency signal strength and transient voltage change rate; Step 4: The visual analysis chip (343) extracts visual features, including hotspot features and discharge spot features; Step 5: The edge data analysis chip (342) uses a weighted decision fusion algorithm to calculate the probability P of partial discharge state occurrence based on electrical signal features and visual features, and sets a threshold for the probability of partial discharge state occurrence. If judge If a partial discharge state is detected, the edge data analysis chip (342) communicates with the central server (4) through the edge communication module (37) to exchange the partial discharge signal and visual data corresponding to the partial discharge state. In step 5, the formula for calculating the probability P of the partial discharge state is: ； in These are weighting coefficients. These are the normalized electrical characteristic values. It is the detection probability of the discharge spot characteristics; Detection probability of discharge spot features The calculation formula is: ； in It is a visible light image frame. Represents the CNN model. It is a thermal infrared temperature matrix. yes The function maps the CNN output to the interval [0, 1], representing the confidence level of the existence of the discharge spot.

2. The edge computing-driven real-time monitoring and diagnosis method for partial discharge according to claim 1, characterized in that: The edge computing-driven partial discharge real-time monitoring and diagnosis system includes a partial discharge monitoring terminal planar moving device (1), multiple device partial discharge edge monitoring terminals (3) and a central server (4). The multiple device partial discharge edge monitoring terminals (3) are detachably installed at the bottom of multiple planar electric sliders (2) of the partial discharge monitoring terminal planar moving device (1), and follow the multiple planar electric sliders (2) to move locally on the top of multiple devices to be tested (101). The device partial discharge edge monitoring terminal (3) monitors the partial discharge signal and visual data of the device to be tested (101), and analyzes whether the device to be tested (101) has a partial discharge state based on the partial discharge signal and visual data. If it is determined that a partial discharge state exists, it interacts with the central server (4) to provide the corresponding partial discharge signal and visual data. If the device partial discharge edge monitoring terminal (3) determines that no partial discharge state exists, it stores the partial discharge signal and visual data at the edge for future reference.

3. The edge computing-driven real-time monitoring and diagnosis method for partial discharge according to claim 2, characterized in that: The partial discharge monitoring terminal planar movement device (1) includes multiple planar tracks (11), multiple planar electric sliders (2) and multiple electric lifting mechanisms (5). The multiple planar tracks (11) are respectively installed on the top of the room (100) where multiple devices to be tested (101) are located. The multiple planar electric sliders (2) are respectively installed on the multiple planar tracks (11). Each planar electric slider (2) moves locally within the area of ​​the corresponding planar track (11). The bases of the multiple electric lifting mechanisms (5) are respectively installed on the multiple planar electric sliders (2). The lifting parts of the multiple electric lifting mechanisms (5) are vertically downward. The multiple device partial discharge edge monitoring terminals (3) are respectively detachably installed at the end of the lifting part of the electric lifting mechanism (5). The device partial discharge edge monitoring terminals (3) move within the top space of the device to be tested (101) under the driving action of the planar electric sliders (2) and the electric lifting mechanisms (5).

4. The edge computing-driven real-time monitoring and diagnosis method for partial discharge according to claim 3, characterized in that: The device partial discharge edge monitoring terminal (3) includes a housing, a multimodal sensor module (32), a visible light visual monitoring module (33), an edge computing module (34), a positioning module (35), a memory (36), and an edge communication module (37). One side of the housing is detachably installed at the end of the lifting part of the electric lifting mechanism (5). The multimodal sensor module (32) and the visible light visual monitoring module (33) are respectively installed inside the housing, and the sensor head is located on the other side of the housing facing the device to be tested (101). The multimodal sensor module (32) monitors pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, and transient voltage signals, and the visible light visual monitoring module (33) monitors visual images. The data output terminals of the multimodal sensor module (32) and the visible light visual monitoring module (33) are respectively connected to the edge computing module (34). The edge computing module (34) calculates whether there is a partial discharge state based on the pulse current signal, mechanical vibration wave signal, electromagnetic wave signal, transient voltage signal and visual image. The positioning module (35) collects the location data of the current device partial discharge edge monitoring terminal (3). The multimodal sensor module (32), the visible light visual monitoring module (33), the edge computing module (34) and the positioning module (35) are respectively connected to the memory (36). The memory (36) is connected to the central server (4) through the edge communication module (37).

5. The edge computing-driven real-time monitoring and diagnosis method for partial discharge according to claim 4, characterized in that: The multimodal sensor module (32) includes a high-frequency current sensor (321), an ultrasonic sensor (322), an ultra-high frequency sensor (323), and a transient voltage sensor (324). The signal output terminals of the high-frequency current sensor (321), ultrasonic sensor (322), ultra-high frequency sensor (323), and transient voltage sensor (324) are respectively connected to the edge computing module (34) for communication, installed in the housing, and the sensing head faces the device to be tested (101).

6. The edge computing-driven real-time monitoring and diagnosis method for partial discharge according to claim 5, characterized in that: The visible light visual monitoring module (33) includes a cover, a visual camera (332) and a thermal infrared sensor (333). The cover is installed inside the housing and faces the device to be tested (101). The visual camera (332) and the thermal infrared sensor (333) are installed inside the housing and located in the middle of the cover. The sensing heads of the visual camera (332) and the thermal infrared sensor (333) face the device to be tested (101) and are respectively connected to the edge computing module (34) for communication.

7. The edge computing-driven real-time monitoring and diagnosis method for partial discharge according to claim 6, characterized in that: The edge computing module (34) includes an edge memory (341), an edge data analysis chip (342), and a vision analysis chip (343). The edge memory (341) is connected to a high-frequency current sensor (321), an ultrasonic sensor (322), an ultra-high frequency sensor (323), a transient voltage sensor (324), a vision camera (332), and a thermal infrared sensor (333) to collect pulse current signals, mechanical vibration wave signals, electromagnetic wave signals, transient voltage signals, and visual images. The edge data analysis chip (342) and the vision analysis chip (343) are connected to the edge memory (341) and the edge data analysis chip (342) is connected to the vision analysis chip (343) to calculate whether a partial discharge state has occurred based on sensor data and vision analysis results. The edge data analysis chip (342) is connected to the central server (4) through the edge communication module (37).

8. The edge computing-driven real-time monitoring and diagnosis method for partial discharge according to claim 7, characterized in that: The central server (4) includes a central communication module (41) and a computer (42), wherein the computer (42) is connected to the edge communication module (37) through the central communication module (41).

9. The edge computing-driven real-time monitoring and diagnosis method for partial discharge according to claim 8, characterized in that: In step 1, the distance is set to 100mm-500mm from the surface of the device to be monitored (101).

Images